基于关系矩阵的工作流日志重复任务识别算法

Detecting duplicate tasks from workflow logs based on casual matrix

针对传统工作流模型挖掘算法不考虑模型中重复任务的存在,导致挖掘出的模型精确度不高的问题,提出一种基于关系矩阵的重复任务识别方法。通过分析工作流执行日志得到所有事件的前驱后继关系,根据不同的模型结构进行事件重命名,再基于同类别重复事件之间的相似度对重复事件进行聚类得到最优识别结果。实验表明,该方法能正确有效地识别工作流日志中的重复任务,减少模型中的不可见任务,最终提高工作流模型挖掘方法的精确度和可理解性。

Aiming at the problem that the traditional process mining algorithms neglected the existence of duplicate tasks in workflow model, which led to the poor precision while using workflow logs produced by these models, a Split and Cluster （SaC） algorithm to detect duplicate tasks from workflow logs based on casual matrix was proposed. The algorithm obtained the casual matrix by analyzing the predecessors and successors of each event in workflow log, the events were relabeled according to specific model structure, and then the optimal result was obtained by clustering the repeating events based on the similarity of each duplicate event. Experimental results showed that SaC could detect duplicate tasks in log correctly and efficiently and reduce the invisible tasks, thus improved the precision and comprehensibility of workflow models.